Canadian Mineral Processors Examine the Fine Art of Grinding

Insights and clues to unravelling these mysteries were at the top of the list during the 33rd Annual Canadian Mineral Processors Meeting last January in Ottawa. Twelve of the 42 papers on the program were devoted to the fine art of grinding. Here are some of the leading edge ideas that were discussed.

Measuring What’s in the Mill

Barrick Gold‘s Holt-McDermott mine in Kirkland Lake, Ont., poured its one-millionth ounce of gold last November. It also marked the completion of a mill expansion. At the same time, a continuous method of measuring the charge volume in the semi-autogenous grinding (SAG) was developed.

Operators had identified a problem, in that the best SAG throughput was achieved just before overload conditions. Consequently, they tended to run below the optimum for fear of overloading the mill.

The solution was to adapt a continuous charge measurement (CCM) system. The CCM system was created in the late 1970s at Skega, which is now Svedala Mill Lining Division. It was adapted and installed at Holt-McDermott in October 1999.

The system measures two data streams. The first is from a strain gauge embedded in a rubber lifter bar to measure deflection. The strain gauge signal is converted into a frequency pulse signal, which is transmitted to a receiver via an antenna wrapped around the mill. The second data stream is from a steel plate mounted on the mill and a proximity switch that counts rotations. A receiver converts both types of data to a 4-20 mA signal and sends them to a computer. Software produces a deflection profile for a full rotation using data from both sources. Then the volumetric load and dynamic angle of repose are determined.

CCM results are shown several ways on a single computer screen (see illustration). There are separate charts for average profile over a number of consecutive cycles and the signal standard deviation, for charge level, and for toe, shoulder and charge angles. Calculated values are shown, such as charge level, charge angle, shoulder angle, toe angle, mill speed, average lifter deflection, and most importantly the peak level and plateau level. There is also a visual representation of the charge.

The CCM solution has been a boon at Hold-McDermott. It estimates the SAG mill static load with accuracy and precision sufficient for process control. Last summer the CCM system was integrated into the mill’s expert system to replace the old strategy of using bearing pressure to control the SAG load. The hardware has proven reliable with the exception of a couple of sensor failures. Moreover, the CCM trajectory looks like it has potential related to mill optimization and condition monitoring.

Grinding Simulation for the Smaller Plant

Simulation has a role to play in grinding evaluation and optimization in plants of every size. For St. Andrew Goldfields‘ 1,300-ton-per-day plant in Matheson, Ont., a simple and elegant simulation model has been developed with the help of Laurentian University of Sudbury, Ont.

The Stock mill treats both the company’s and custom ores. It has little instrumentation, which makes optimum performance a challenge. The grinding circuit consists of a primary ball mill and two regrind ball mills. It is monitored mostly by manual means, and adjustments are made to pulp densities in various streams.

The new simulator is based on the ubiquitous Microsoft Excel spreadsheet and its built-in Visual Basic programming tool. The program creates empirical simulation by correlating dependent and independent process variables. The correlation is established through multiple linear regression analysis of historical plant data. Product size predictions and mass flow rates can then be made for a specific set of operating conditions. Optimization is obtained by identifying the conditions that obtain desired product sizes. Material balance calculations are also possible.

The simulator functions are launched from a custom toolbar and macro buttons in the Excel program. On-screen menus guide the user through selection of various input data (sieve sizes, raw data sets, pulp densities, fresh feed characteristics, etc.). Simulated results for the grinding mills are viewed by clicking buttons on their discharge stream icons. Clicking other buttons returns the R-square values for the regressions or the minimum/maximum values of process variables. Details of a particular unit in the flowsheet are obtained by clicking on the unit icon. Simulations are stored for later review, either individually on the flowsheet or collectively in a spreadsheet.

Operators at the Stock mill are pleased with the Excel-based simulator. It has proven to be an easy-to-use, fully functional solution. With it, they found that ore throughput, ore hardness, primary cyclone overflow pulp density and secondary cyclone underflow pulp density were the most influential factors in determining the final grind size in their plant.

Simulation at a Large Plant

Barrick Gold has expanded the grinding circuit at its Goldstrike mill in Nevada to cope with both increasing ore hardness and the need for finer grinding. There are two grinding lines, each having a SAG mill, pebble crusher and ball mill. The decision to add another ball mill to one circuit was made in part using a comminution economic evaluation tool (CEET), which was introduced at the 1999 CMP meeting.

CEET is made up of three primary components: a ball mill model based on Bond’s third theory of comminution; a SAG/AG model using the SAG power index (SPI) energy relationship; and a data set of ore hardness. At Goldstrike, laboratory values for Bond work index and SPI were distributed across the Betze mine block model. That data set was then analyzed with CEET to evaluate and design SAG and AG circuits. Since CEET is a new tool, investigators also conducted parallel circuit sizing using conventional methods for comparison.

After all the testing, Barrick chose to add a 3,200-hp regrind ball mill to the Line 1 SAG/ball mill grinding circuit. The sizing of the mill was primarily done using conventional techniques, but CEET tests were run to ensure that it would achieve the desired efficiency on a block-by-block basis. CEET also made possible performance comparisons had no expansion been implemented.

CEET was commissioned at Goldstrike only in January 2000, but it appears to have potential for further refinement and optimization of the grinding circuits. The model has been calibrated to actual operating conditions and achieved very realistic simulation results. It has proven its ability to compare Bond and SPI methods in rating SAG and ball mill capacities and, hence, to compare the limitations of each.

More on Bond and SPI

The value of using the Bond work index and SPI at Kinross‘ remote Kubaka gold project in Russia was reported by John Starkey of Starkey & Associates. As the temperature in Russia goes down, the need to keep heated building space to a minimum goes up. In-circuit crushing and external conveyors were not wanted. Energy and grinding efficiency for both the SAG and ball mills were foremost in design considerations.

A large number of ore samples were tested for hardness using the Bond work index and the MacPherson autogenous work index. Gross pinion power necessary to grind to specific sizes was estimated from the results. Electrical and mechanical losses in the motor or drive train, which can range from 2% to 10%, were figured in at 8%.

When using SPI to calculate grinding energy consumption, the calculation differs slightly. Gross SAG mill power is measured at a meter and includes the drive train power loss. This is a practical method of measurement because most SAG mills have a synchronous motor coupled directly to the pinion. Electrical and mechanical efficiency loss is only about 6% in such installations. The Kubaka grinding circuit was designed using SPI calibration work (originally presented at the SAG ’96 conference), perhaps the first such installation to be planned using these calculations.

O
ne key to using SPI during the design of a grinding circuit, is to use the value of the hardest ore, not an average hardness. With that caveat in mind, engineers did their tests and chose the mills for the Kubaka project. They selected a 20-ft diameter by 7-ft-6-in SAG mill, which is close to the original design choice. The ball mill is 13-ft-6-in diameter by 18-ft, and was originally intended for a rod mill application. It has a large diameter trunnion and an oversized motor. To bring the ball mill speed up to 75% of critical from 66%, the 24-tooth pinion was replaced with a 27-tooth one.

Equipment selection using SPI data was deemed a success. The mill met or exceeded design tonnage since start up in 1997. Comparative data for the period March-August 2000 is given in Table 1. The excellent operating results are due in part to slightly finer feed, coarser final product, and designing for the hardest ore to be treated.

Advances in Dry Grinding

Dry grinding, despite its higher cost than wet grinding, is sometimes the only option. Such is the case at Barrick Gold’s Goldstrike mill. Almost half of the outlined orebody contains material unsuitable for autoclave treatment because it contains high levels of total carbonaceous material. To treat this “double refractory” ore, an oxygenated roasting plant has been built. However, the material roasts at a temperature too low to efficiently drive off water, were the ore to be wet ground.

The dry grinding plant has a daily capacity of 12,000 tons and an availability record of 90%. It consists of a single Svedala 500-hp gyratory crusher, an 8-ft by 24-ft Simplicity vibrating screen, a Nordberg 800-hp cone crusher, and two Krupp-Polysius double-rotator grinding mills. Each one is 19-ft-diam by 29-ft-long and has a variable-speed gearless motor drive. Such mills are commonly used in the cement industry but only rarely for mineral processing.

Crushed ore and hot gas are fed into the drying chamber of each mill where the moisture content is reduced to 0.5%. Dry ore passes through a diaphragm into the coarse grinding chamber where a charge of 4-inch chrome-steel balls reduces the size to -1/4-inch. A second diaphragm permits smaller particles to exit the coarse grinding chamber. The mills are air-swept, forcing small particles through a static classifier mounted above the second diaphragm. The particles then pass through a baghouse for collection.

The collected particles are sent either directly to the roaster feed bins or to dynamic classifiers. Here the stream is sized once more, with the coarser material directed to the feed end of the drying chamber. The finer portion reports to the fine grinding chamber of the mill, which is charged with a mix of 2-inch and 21/2-inch balls. The fine grinding chamber is also air-swept, again forcing the fine particles through the second diaphragm near the middle of the mill and upwards to the static classifier.

Despite the complexity of a dry grinding circuit compared to a wet one, Goldstrike employees got their new facility up and running easily. The first mill was fed with waste rock in February 2000 and run-in with a reduced ball charge. Two days later it was operating with a full charge at desired capacity. Within an hour of turning over the second mill, it was also running at the designed capacity of 278 tons per hour and a 28% steel charge.

Grinding is possibly the most important procedure in mineral processing. Yet it remains difficult to understand and control. So many questions need answers: What is really going on in the mill? How can liberation be modelled? optimized? automated? The solutions have huge money-saving potential, particularly in energy and supplies.

Table 1: Actual and design data at the Kubaka gold mill

Item	Actual	Design
Average SAG	103.2	81.0 feed (t/h)
Gross SAG specific	10.7	13.8 energy (kWh/t)
Gross Ball Mill	11.0	16.9 specific energy (kWh/t)
Calculated	12.0	15.5 Bond work index (kWh/t)
SAG operating	96.1	90 availability (%)

---